[Info-vax] Reimplementing VMS, was: Re: HP adds OpenVMS Mature Product Support beyond the end of Standard Support

[Stephen Hoffman]

On 2014-02-04 03:44:16 +0000, JF Mezei said:

> Before 8086 gained quickpath/CSI, it had poor multi-core/multiprocessor 
> scalability, something where IA64 had an advantage, and something which 
> the BCS customer base looked for for large "mainframes".

With a twelve-core Xeon processor, you've a fairly big server in one socket.

With a per-core licensing scheme, you've a pretty big bill for some of 
the software products, too.

If recent history is at all predictive, the core counts and the 
component integration are only going to increase.

While QPI definitely helps when you're scaling past one socket, VMS 
already has support for NUMA box designs, and that'll be useful even 
with QPI.

Yes, QPI (or pretty much anything faster than the old FSB) also helps 
with getting I/O on and off the processor.

AFAIK and FWIW, the existing VMS SMP kernel won't scale past a 
hyperthreads-capable sixteen-core dual-socket box in its present design.

Until you can find a revenue model that would sustain an operating 
system project, and until you can secure the rights to all of the 
necessary source code, and a whole host of other non-technical details, 
the whole discussion of QPI and EFI and the rest is moot and merely 
serves as entertainment — and if you can find that revenue model, then 
you could potentially suggest it to HP, as they might decide to make a 
go of it.


-- 
Pure Personal Opinion | HoffmanLabs LLC